AI Agents in Games: From Theory to Practice at IVS2026
The session frames EVE Online as the ultimate benchmark for AI agency, while acknowledging that practical, cost-effective LLM agents in games remain distant.
Reporting from 1 source: 4Gamer.net.
At IVS2026, a panel of investors and entrepreneurs from gaming, healthcare, and venture capital discussed the current state and future of AI agents. Topics included the lineage from AlphaGo to potential AI in EVE Online, the gap between sophisticated bots and LLM-based agents, and the challenge of making agents economically viable in live games.
On July 1, 2026, at the startup conference IVS2026, a panel titled 'How Frontier Models, Consumer Scale, and Strategic Capital are Activating Real-World AI Agents' brought together four figures from different industries. Jordan Fisher (Antler/Zehitomo), Justin Waldron (Zynga/Playable Intelligence), Hilmar Veigar Petursson (Fenris Creations), and Amir Dan Rubin (Healthier Capital) discussed the current state of AI agents.
Petursson, whose team created EVE Online, described the game as a world where players exercise agency, making AI agents a natural fit. He recalled a decade-old conversation with DeepMind founder Demis Hassabis about creating 'AlphaEVE,' which they agreed would be the ultimate AGI benchmark. However, Petursson noted that while bots have existed in MMOs for decades, running LLM-based agents in EVE Online currently harms ROI, and DeepMind's SIMA system has an advantage by operating directly from pixel data.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.